Predicting and containing epidemic risk using on-line friendship networks

Area of Science:

• Epidemiology
• Network Science
• Computational Social Science

Background:

• Infections spread through human encounters, but detailed data collection is challenging.
• Online social networks offer a potential proxy for physical interactions.
• Predicting and containing epidemic risk requires understanding transmission dynamics.

Purpose of the Study:

• To evaluate the utility of online social networks in predicting and containing epidemic risk.
• To compare the predictive accuracy of friendship networks versus encounter networks.
• To assess strategies for epidemic containment using online social network data.

Main Methods:

• Constructed a time-varying encounter network and a static friendship network from user data.
• Employed computer simulations to model stochastic infection processes on both networks.
• Compared prediction accuracy and containment strategies against a benchmark of known encounters.

Main Results:

• Friendship networks identify at-risk individuals but with lower accuracy than encounter data.
• Static encounter networks outperform static friendship networks in risk prediction.
• Periodical monitoring of encounter networks corrects friendship network predictions, achieving high accuracy.
• Friendship networks provide valuable information for cost-effective epidemic containment strategies.

Conclusions:

• Online social networks, particularly when combined with encounter data monitoring, can effectively predict and help contain epidemic spread.
• Friendship network data is valuable for targeted immunization strategies, even with limited budgets.
• Leveraging online social network data offers a feasible approach to managing public health risks associated with infectious diseases.